Electrochemical plating is a process of depositing a metal layer on a metallic or non-metallic substrate. The technology is used in a variety of industrial applications including integrated circuit fabrication, semiconductor packaging, printed circuit board manufacturing, metallic coating and finishing, and others. In an electroplating process, an electric current is passed through an electroplating cell comprised of a working electrode (cathode), counter electrode (anode), and an aqueous electrolyte solution of positive ions of the metal or metals to be plated on a substrate in physical contact with the cathode. By applying a potential to the electrodes, an electrochemical process is initiated wherein cations migrate to the cathode and anions migrate to the anode. Metallic ions deposit on a substrate attached to a cathode to form a metal coating.
In semiconductor applications, the plating of metals such as Au, Sn, Cu, and Pb is used for packaging and printed circuit board products. In particular, a dual damascene Cu plating method is commonly employed to fabricate metal lines for integrated circuits. Typically, a seed layer is sputter deposited in openings to line the sidewalls and bottom surfaces of vias and trenches. Then, an electroplating process fills the openings, preferably in a direction from bottom to top to avoid formation of pin holes or voids that can degrade device performance.
U.S. Pat. No. 6,156,167 discloses an apparatus for electrochemically treating semiconductor wafers and U.S. Pat. No. 6,159,354 describes an electric potential shaping method for electroplating that can be implemented in a dual damascene copper process. In addition, U.S. Pat. No. 5,985,126 describes a semiconductor workpiece holder used in electroplating systems. Conventional manufacturing processes generally comprise at least a plating cell, post-plating module, wafer handling device, wafer transfer mechanism, plating bath, and a chemical monitoring device and dosing system.
In addition to inorganic constituents in the form of metal salts, plating baths contain organic additives required for achieving the desired deposition properties. Types of organic additives include a suppressor otherwise known as a carrier or wetting agent, a brightener also known as an accelerator, grain refiner, starter, etc., and a leveler also referred to as a leveling agent, momentum deposition reducer, etc.
A consequence of the electroplating operation is degradation of the organic species over time. As additives degrade, their concentrations deviate from the nominal (or target) value. Moreover, the accumulation of degradation products may adversely affect the properties of the deposited metal film. For example, carbon, nitrogen and sulfur may be incorporated in a Cu film and thereby lower electrical conductivity.
Several methods have been proposed to maintain the consistency of organic additives in a copper plating bath. U.S. Pat. No. 6,458,262 describes a method of removing organics from the plating bath and analyzing the reconstituted organic fraction with high pressure liquid chromatography (HPLC). The deviation of the measured result from the additive target concentration is the amount to be replenished.
U.S. Pat. No. 6,471,845 describes a Smart Dosing method to compensate for consumption of additives in a plating bath. A bath maintenance feature mathematically predicts concentration changes in additive species based on the assumption that a given additive may degrade linearly due to the passage of time and charge. The predictive algorithm compares the predicted concentration of a species to its nominal concentration and calculates a quantity of the additive from a dosing reservoir that would be required to reset the bath. A metering pump for that additive species is then activated and the required dose is administered to a central bath reservoir. However, Smart Dosing does not perfectly compensate for additive consumption. As a result, actual analysis of batch constituents should be performed periodically. If measured concentrations are below the target values, additives will be added as corrective dosing.
Corrective dosing requires analytical instruments and methods to quantify the amount of additives in a plating bath. U.S. Pat. No. 5,192,403 relates to a cyclic voltammetry scan (CVS) method for measuring concentration of a plating bath component. U.S. Pat. No. 7,270,733 discloses a method involving chemometric analysis of voltammetric data and is commonly referred to as RTA (Real Time Analyzer). U.S. Pat. No. 7,531,134 describes an apparatus utilizing isotopically labeled spikes, an electrospray ionizer, and mass spectrometry to characterize concentrations of plating solution constituents.
As a plating bath ages during usage, the amount of degradation products increases and it is generally recognized that contamination from atoms such as carbon, nitrogen, and sulfur in deposited metal films rises as a result. Reduction of the contamination is accomplished by “bleed-and-feed” in which a certain fraction of plating solution is dumped and replaced with fresh materials. U.S. Pat. No. 6,471,845 points out that with the bleed-and-feed approach degradation products can reach limiting (or steady state) concentrations where they will not degrade the process. In U.S. Pat. No. 6,827,832, an electrochemical process is disclosed to break down and remove degradation compounds and refresh additives in a plating bath.
As for the additives, the suppressor is typically a polyethylene glycol (PEG) or a block copolymer of polyethylene oxide (EO) and polypropylene oxide (PO) and it absorbs on a copper cathode aided by the chloride present in the bath solution. The suppressor functions as an inhibitor of field deposition to facilitate bottom-up and void-free fill in dual damascene Cu plating and through silicon via, and also serves as a surface wetting agent to reduce deposition defects caused by lack of contact between plating solution and substrate. The suppressor breaks down into lower molecular weight fragments during wafer processing.
The accelerator is typically a sulfur containing organic species such as thiourea, cystine, 2-mercaptoethylsulfonate, 3-mercaptopropylsulfonate, and dimers of some sulfur derivatives. Accelerator functions include promoting bottom-up fill in dual damascene interconnects and through silicon via, and refining deposited metal grain structure. An accelerator is likely to be oxidized during plating operations to sulfones, sulfoxides, sulfonates, and other products with higher oxidative states.
A leveler is usually a nitrogen containing polymer that reduces momentum deposition over trench, via, and recessed areas on a substrate. A leveler may undergo reductive or oxidative reactions during plating. Degradation products of suppressor, accelerator, and leveler may possess their own electro activity and thereby influence an electroplating process.
One shortcoming of electrochemical analytical techniques such as CVS and RTA is that they are not very selective. CVS and RTA measure quantities related to charges that pass through analytical electrodes under analytical conditions but such charges are not exclusively dependent on concentrations of additives being measured and can be merely a combined effect of all bath constituents. Therefore, additive concentrations detected by CVS and RTA are not true additive concentrations in an aged plating bath where degradation products exist. Instead, they represent the combined analytical electro effect of additives and degradation products expressed in terms of concentrations of pure additives. Referring to FIG. 1, suppose an actual bath solution contains 5 mg/liter additive and 10 mg/liter degradation products while a hypothetical bath contains 5 mg/liter plus 3 mg/liter additive and no degradation products. Both of the hypothetical and actual baths have the same measured response on the two electrochemical analyzers, respectively. CVS or RTA then reports out additive concentration in the hypothetical bath (5+3=8 mg/liter) as a representation of the actual bath additive composition. CVS and RTA further assume that the hypothetical bath would have the same process output during wafer processing on a plating tool as the actual bath because they have the same analytical response on the analyzers. A chemical process control system for a plating bath using RTA is disclosed in U.S. Patent Application 2006/0172427.
Electrochemical response is a complex process and involves variables related to surface activity, substrate type, molecular structures of surface active materials, mass transport, redox potentials and activation energies, etc. Actually, elements of an electroplating process are very different from those of CVS or RTA. Further, elements of CVS are different from those of RTA. For example, the electrode substrate in CVS and RTA is Pt while in copper plating it is Cu. In CVS and RTA, the electrode is flat, and also rotates in the case of CVS. In dual damascene copper plating, the electrode surface is blanket copper film over topography that includes vias, trenches and recesses. Current density and electrode potentials in analysis and wafer processing are not equivalent. Wafer processing typically uses a galvanostatic approach while CVS and RTA take a potentiostatic path. U.S. Pat. No. 7,022,212 shows an analytical method to simulate actual conditions on a wafer and to measure additive concentration and mass transfer of plating components to control a plating bath composition.
CVS and RTA analytical response is an average over localities across an electrode surface. On the other hand, in Cu dual damascene, process response such as bottom-up fill is a local event driven by accumulation of accelerator at the bottom of features due to surface area reduction as plating progresses according to one theory. Bottom-up fill is also a function of the intrinsic ability of accelerator to displace suppressor adsorbed on a copper surface unlike CVS and RTA where suppressor is displaced from platinum.
Clearly, it is questionable and an oversimplification to assume that degradation products which produce a CVS or RTA analytical response equivalent to a certain amount (X) of fresh accelerator (or other additive) will also generate the same response as an X amount of fresh accelerator in a plating process. In fact, it is extremely unlikely that this relationship would occur in a large variety of electroplating operations practiced in the industry. Moreover, there is plenty of evidence to support the opposite conclusion. For instance, CVS results in an aged bath do not agree with those from RTA. This outcome can be caused by a variation between CVS and RTA analytical conditions which then leads to a different analytical electro response for a given amount of degradation products.
Although selectivity of CVS and RTA was improved in recent years to a point where CVS and RTA are now the preferred choices for bath monitoring in the plating industry, selectivity improvement is still limited by two fundamental realities. One is that with an inherently poor selectivity methodology and increasing number of active components (e.g. degradation products) present in solution, there is a limitation to how well interference of one additive by other constituents can be separated. Secondly, separation of interference in electrochemical analysis requires availability of pure materials for analytical method development. For example, pure samples of accelerator degradation products are needed to exclude their effect from the total CVS or RTA sensor response so that accelerator only response can be derived for determining accelerator concentration. However, those degradation products are rarely identified and cannot be sourced in most cases. Therefore, the analytical result from CVS or RTA relative to suppressor, accelerator, or leveler is in fact the total analytical electro response from the species being analyzed and degradation products that is exhibited during analysis, although the response is expressed in terms of concentration of pure additive that would have produced the same analytical electro activity under the same analytical condition in the absence of degradation products.
Chromatography methods have been applied to analyze organic additives as mentioned by B. Newton et al. in “Analysis of copper plating baths suppressors and levelers”, Proc. Electrochem. Soc., V2000-27, page 1, Dec., 2000, by R. Palmans et al. in “Ion-pair chromatography of bis(sodium-sulfopropyl)disulfide brightener in acidic copper plating baths”, Journal of Chromatography A, 1085, pp. 147-154 (2005), and by K. Hong et al. in “A new metrology system of organic additives in copper electroplating baths”, Journal of the Korean Physical Soc., Vol. 43, No. 2, pp. 286-289, Aug., 2003.
Mass spectrometry was also reported to be applicable to plating bath analysis by R. Mc Donald in “Automated mass spectrometry to detect impurities in harsh acid chemistries”, Solid State Tech., Vol. 49, Issue 6, Jun., 2006.
Arguments have been made by established industry participants that conventional quantitative analytical chemistry techniques such as HPLC and mass spectrometry are not appropriate choices for monitoring a plating bath because they do not include electro activity of degradation products in reported additive concentrations which are used as input variables to control a plating bath process. As a result, analytical technology based on HPLC and mass spectrometry which is popular in chemical, pharmaceutical, and biotechnology industries has been used sparingly in the electroplating industry. However, some chemical species resulting from additive degradation in plating baths do possess electro activity and could have a significant impact on the electrochemical response of a plating bath. For example, one of the common suppressing agents, high molecular weight polyethylene glycol (PEG) breaks down into low molecular weight PEG which is known to have CVS activity according to U.S. Pat. No. 6,749,739. Also, 3-Mercaptopropyl sulfonic acid (MPSA) is produced electrolytically as a degradation product of bis(sodiumsulfopropyl)disulfide (SPS), an accelerator. U.S. Pat. No. 7,291,253 indicates a CVS activity for MPSA higher than that of SPS under the same analytical condition.
A current practice in the electroplating industry is to rely on the bleed-and-feed approach where usually 10% to 30% of a bath reservoir is drained and replaced by fresh solution each day to maintain degradation products at a steady state level. Therefore, plating baths are expected to have consistent process performance over time under bleed-and-feed conditions when bath additive concentrations are supposed to be kept at a constant level. Since current electroplating processes are known to produce unexpected results when a bath becomes aged and the only solution in such an event is to dump the bath, this occurrence suggests that certain elements of the bleed-and-feed process chemistry are not controlled and understood.
Referring to FIG. 2, the amount of degraded additive present in a bath after a certain number of processing days is expressed in terms of a % of the nominal concentration. The plots for bleed/feed rates of 10%, 20%, and 30% assume a production environment including 60 wafers/hour throughput, 70% utilization, and 90% uptime. The amount of degradation products existing under steady state in copper plating processes range from 0 to 1.6 times the target additive concentration. Within this concentration range, the electro activity of degradation products probably represents a small portion of total electro activity and its rise and fall due to fluctuation in degradation product content over time is tolerated by the process. In the Palmans reference cited previously, results showed that additive concentration in a bath presumably kept stable by electrochemical analytical techniques had approximately ±50% variation from an average value. Although significant, such concentration variation may be tolerated by the process.
Curves 20, 21, 22 in FIG. 2 indicate an increasing amount of degradation products with reduced bleed-and-feed. This condition can thus create a bath with sufficient degradation products to cause a rise and fall in additive concentration wherein the fluctuation leads to process variability. In other words, a bath kept at constant total analytical electro response may see sufficiently large additive concentration fluctuation to impact deposition because of the variation in degradation composition associated with reduced bleed-and-feed. Consequently, it is likely that unexpected and unexplainable process outliers will become more frequent if the bleed-and-feed rate is reduced. In summary, the necessity of a 10%-30% daily bleed-and-feed and associated material related cost may be precipitated by using non-selective analytical electro activities as input variables for bath chemistry control.
Unfortunately, current control methodology for electroplating processes that rely only on electrochemical analytical techniques are not sufficiently reliable to reduce dependence on the bleed-and-feed approach which costs billions of dollars worldwide because of the expense incurred with a high volume of replenished components and frequent disposal of up to 10% to 30% of the plating solution. Furthermore, current production processes are still subject to unexpected and/or unexplained performance failures even with in-control bath chemistry. The only solution is to dump and renew the entire bath which drives material cost higher and presents a process reliability issue. Therefore, an improved electroplating process control system and methodology is needed to reduce production cost and improve reliability.